LIME

Bases: MultitaskExplainerMixin, AttributionExplainer

Local Interpretable Model-agnostic Explanations (LIME) is a perturbation‑based approach that explains individual predictions by fitting a simple, interpretable surrogate model locally around the prediction of interest. By sampling perturbed versions of the input and weighting them by their proximity to the original instance, LIME learns per‑feature importance scores that approximate the behaviour of the underlying black‑box model in that local region.

Reference: Ribeiro et al. (2016). "Why Should I Trust You?": Explaining the Predictions of Any Classifier. Paper

Examples:

>>> from interpreto import Granularity, Lime
>>> from interpreto.attributions import InferenceModes
>>> method = Lime(model, tokenizer, batch_size=4,
>>>               inference_mode=InferenceModes.LOG_SOFTMAX,
>>>               n_perturbations=20,
>>>               granularity=Granularity.WORD,
>>>               distance_function=Lime.distance_functions.HAMMING)
>>> explanations = method(text)

Parameters:

Name	Type	Description	Default
model	PreTrainedModel	model to explain	required
tokenizer	PreTrainedTokenizer	Hugging Face tokenizer associated with the model	required
batch_size	int	batch size for the attribution method	4
granularity	Granularity	granularity level of the perturbations (token, word, sentence, etc.)	WORD
inference_mode	Callable[[Tensor], Tensor]	The mode used for inference. It can be either one of LOGITS, SOFTMAX, or LOG_SOFTMAX. Use InferenceModes to choose the appropriate mode.	LOGITS
n_perturbations	int	the number of perturbations to generate.	100
perturb_probability	float	probability of perturbation.	0.5
distance_function	DistancesFromMaskProtocol	distance function used to compute weights of perturbed samples in the linear model training.	COSINE
kernel_width	float | Callable | None	kernel width used in the similarity_kernel. If None, the kernel width is computed using the default_kernel_width_fn function.	None
device	device	device on which the attribution method will be run	None